Risk of Breast Cancer in Relation to the Use of Injectable Progestogen Contraceptives and Combined Estrogen/Progestogen Contraceptives

Transcription

1 American Journal of Epidemiology Copyright by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol., No. Printed in U.S.A. Risk of Breast Cancer in Relation to the Use of Injectable Progestogen Contraceptives and Combined Estrogen/Progestogen Contraceptives Samuel Shapiro, Lynn Rosenberg, Margaret Hoffman, Hanneke Truter, Diane Cooper, Sowmya Rao, David Dent, Anne Gudgeon, Jakobus van Zyl, Judith Katzenellenbogen, and Ross Bailie Recent studies have suggested that progestogen-only contraceptives and combined estrogen/progestogen oral contraceptives (COCs) may increase the risk of breast cancer among women less than years of age or among recent users. The authors conducted a case-control study, in which cases of breast cancer (n = ) and controls (n =,) hospitalized for conditions unrelated to contraceptive use were interviewed from to 7 in hospitals in greater Cape Town, South Africa. The women were aged - years, resided in a defined area around Cape Town, and were Black or of mixed racial descent. The relative risk for exposure to injectable progestogen contraceptives (IPCs), mostly depot medroxyprogesterone acetate, was (% confidence interval (Cl).7, ). There were no consistent associations within categories of age or recency or duration of use. For exposure to COCs, the overall relative risk was (% Cl.,.). Among women below age years, the relative risk was.7 (% Cl.,.), and it was unrelated to the duration or recency of use. The findings suggest that IPCs do not increase the risk of breast cancer, and that COCs may increase the risk among women below age years, although bias cannot be excluded. Am J Epidemiol ^ :-. breast neoplasms; contraceptives, oral, hormonal Depot medroxyprogesterone acetate (DMPA; trade name, Depo-Provera) is an injectable progestogen contraceptive (IPC) that has been used worldwide by millions of women since its introduction more than three decades ago (). DMPA induces mammary tumors in Received for publication February,, and accepted for publication May,. Abbreviations: Cl, confidence interval; COCs, combined estrogen/progestogen oral contraceptives; DMPA, depot medroxyprogesterone acetate; IPC, injectable progestogen contraceptive;, relative risk; WHO, World Health Organization. Slone Epidemiology Unit, Boston University School of Medicine, Brookline, MA. Department of Community Health, University of Cape Town Medical School, South Africa. Medical Research Council of South Africa, Belleville, South Africa. Department of Surgery, Groote Schuur Hospital and University of Cape Town Medical School, Observatory, Cape, South Africa. Tygerberg Hospital and University of Stellenbosch Medical School, Belleville, South Africa. Reprint requests to Dr. Samuel Shapiro, Slone Epidemiology Unit, 7 Beacon Street, Brookline, MA -. Note: Ethnic groups are classified in this paper according to terms used by the previous Apartheid government of South Africa. The appropriateness of using "race" as a variable in scientific research is currently being debated, both in South Africa and internationally (, ). The issues are complex. While the debate continues, the authors have felt obliged to use the old terms, since there were clear differences in breast cancer incidence, contraceptive usage patterns, and covariates according to ethnic group. In doing so, the authors acknowledge that the system of Apartheid resulted in disease patterns and in access to health care according to a racial classification that was not legitimate. beagle bitches (), and for that reason it was not licensed in the United States for use as a contraceptive until (), after two epidemiologic studies independently reported no evidence of an overall increase in the risk of breast cancer among exposed women (, 7). However, the studies raised the suspicion that DMPA may increase the risk among women below age years, or, alternatively, among currently exposed women, and for a few years after they stop. In contrast to DMPA, combined estrogen/progestogen oral contraceptives (COCs) have been subjected to repeated epidemiologic assessment of breast cancer risk from the time they were introduced (-). The evidence in early studies was generally reassuring. In recent years, however, associations have been observed, again in relation to the occurrence of breast cancer before age years or among currently or recently exposed women. In a collaborative reanalysis of data from studies (, ), the risk was highest among current COC users, with a decline to baseline some years after cessation of use. For the most part, the associations have been of modest size, and it has not been clear whether they were causal or due to bias (-). In South Africa, DMPA has been used more commonly, and for longer durations, than anywhere else in the world (), and recently another injectable progestogen, norethisterone enanthate, has also been used. The high usage rates of IPCs afforded an oppor- Downloaded from by guest on October

2 Hormonal Contraception and Breast Cancer 7 tunity to evaluate breast cancer risk overall and within subgroups. COC use has also been quite common (). In this report, we evaluate the risk of breast cancer among women exposed to IPCs and COCs. MATERIALS AND METHODS Study base A case-control study was conducted in a base population that comprised all Black and Colored women (women of mixed racial descent) who lived within a defined area surrounding Cape Town, and who were - years of age. White women were not included because family planning clinic records indicated that IPC use was uncommon. In (), the study population comprised, women (Colored,,; Black,,7). Cases The cases were women with first occurrences of invasive breast cancer treated between January and October 7 at two tertiary care hospitals in Cape Town. Women with carcinoma-in-situ, those who had a previous history of cancer, or those who had not resided in the study region for at least months out of the preceding months, were excluded. There were eligible cases, of whom three were too ill to be interviewed, and three refused. Among the remaining cases, the median age was years; cases (7 percent) were Colored and ( percent) were Black. Until, virtually all cases in the study region were seen at the two hospitals because these were the sole institutions that provided specialized care, including radiotherapy and chemotherapy. Thereafter, such services also became available elsewhere, and were accessible to women with comprehensive medical insurance (membership in a medical aid society). Based on cases identified in and, the estimated annual incidence of breast cancer was. (Colored,.; Black,.7) per,. Among women aged -, -, and - years, the rates were., 7., and. per,, respectively. Controls Am J Epidemiol Vol., No., The controls were women admitted for diagnoses judged to be independent of contraceptive use and breast cancer risk: women admitted for conditions such as venous thromboembolism, ischemic heart disease, or benign breast disease were not eligible. The controls were frequency-matched to the cases in a ratio of up to : for decade of age, ethnic group, and area of residence. For cases referred from outlying hospitals, controls referred from the same areas were enrolled, or failing that, they were recruited in hospitals located in those areas ( subjects; percent of the total). As with the cases, women with a history of cancer, and those who had not been resident in the study region for at least months out of the preceding months were excluded. There were,7 eligible controls, two of whom refused to participate. Among the remaining, controls, the median age was years;, ( percent) were Colored, and ( percent) were Black. Data collection A questionnaire written in English, Afrikaans, or Xhosa was administered in the subjects' preferred language by qualified nurses. To maximize recall, the interviewers compiled a calendar of significant personal events (age at menarche, births, etc.), and periods of contraceptive use were related to those events. COC packages were shown to the women to facilitate identification. DMPA and norethisterone are administered at - and -month intervals, respectively. Thus, when the women could not remember the name of the IPC, it could be identified either as DMPA or norethisterone from the reported time interval between injections. Data analysis Information on a wide range of variables was recorded in order to control confounding, and relative risks (odds ratios) were estimated using unconditional multiple logistic regression. Family history of breast cancer, history of benign breast disease, late age at first birth, and low parity or nulliparity were associated with breast cancer, but not with IPC or COC use. Breast feeding was not related to breast cancer risk. These factors were nevertheless included in initial multivariate models, but they had no effect on the relative risk estimates for contraceptive exposure; adjustment for urban or rural area of residence also had no effect. High socioeconomic status was strongly associated with breast cancer, and this factor was also related to contraceptive use. Of the indices used to measure socioeconomic status (years of education, employment status, medical insurance coverage), the one that most strongly differentiated between the cases and the controls was medical insurance coverage: no insurance, membership in a sick fund (limited insurance), and membership in a medical aid society (comprehensive insurance) were used to denote low, intermediate, and high status, respectively. The final regression models included terms for IPC and COC use, age, ethnic group, and socioeconomic status, as indicated by insurance coverage. Downloaded from by guest on October

3 Shapiro et al. RESULTS Among the, controls (table ) 7 percent had used IPCs, and 7 percent were exposed for > years. For COCs, the corresponding rates were and percent. There was little variability according to diagnosis. Injectable progestogen contraceptives Among the cases and, controls, ( percent) and, (7 percent), respectively, had used IPCs (relative risk () =, percent confidence interval (CI).7, ) (table ). For total durations of use that extended from < year to > years, the overall relative risk estimates ranged from to.. For intervals since first use that extended from < years to > years, the overall estimates ranged from to. For intervals since last use that extended from to > years, the overall relative risk estimates were or less. The percent confidence intervals for all of the foregoing estimates included.. The relative risk was. ( percent CI,.) among current IPC users (women last exposed less than a year previously). However, of the cases and controls last exposed - years previously (table ), and, respectively, had last received IPCs in the previous - years, and the relative risk estimate was. Among current IPC users the relative risk estimate of. did not vary significantly according to duration of use (data not shown). In the age-specific data (table ), among women who were aged ^ years, the relative risk for current IPC use was. ( percent CI.,.); among women aged < years and - years, the corresponding estimates were and., respectively, and compatible with unity. In the remaining categories of duration, and of intervals since first and last exposure, there were no significant associations, and the relative risk estimates ranged from. to.. There were and, Colored cases and controls, of whom ( percent) and (7 percent), respectively, had used IPCs ( =, percent CI.7, ). Among and Black women, the corresponding numbers were 7 ( percent) and 7 ( percent) ( =., percent CI,.). DMPA was used by 77 cases and, controls ( =, percent CI.7, ); norethisterone was used by cases and controls ( =.7, percent CI.,.). To assess the possibility that COC use may have modified the effects of IPCs, the analysis was repeated after excluding cases and controls exposed to COCs. Among the remaining cases and controls, and, respectively, had used IPCs ( =, percent CI,.). One study has reported an elevated risk lasting up to years after a single injection of DMPA (): in the present study, one case and seven controls had received a single injection (in all instances DMPA) less than years previously (crude =.7). It has also been hypothesized that hormonal exposure during adolescence or early adulthood may increase the risk (7): there were cases and 7 controls who were aged years or younger when they were first exposed to IPCs ( =, percent CI., ). Selection bias could have occurred if contraceptive usage patterns differed between the cases enrolled in this study, and those who were not enrolled because they attended private clinics. There were cases and, controls who were not members of medical aid societies (and who could not attend private clinics), of whom 77 and,7, respectively, were exposed to IPCs ( =., percent CI,.). Bias could also have occurred if otherwise occult breast cancer was selectively diagnosed among IPC users (detection bias) (). There were cases and, controls who had never previously undergone a breast examination, self-examination, or mammography: among TABLE. Use of injectable progestogen contraceptives (IPCs) and combined estrogen/progestogen oral contraceptives (COCs) among, controls in greater Cape Town, South Africa, -7 No. Trauma (n = ) %* Acute infections (n--= ) No. Diagnostic category % Other conditions (n = ) No. % No. Total (n=i. ) %* Downloaded from by guest on October IPCsf Ever > years OCst Ever > years , 7 7 * Percents are rounded off. t Totals exceed % because IPC and COC exposures overlap. Am J Epidemiol Vol., No.,

4 ! lo. ro o TABLE. Relative risks* () (% confidence intervals (Cl)) for the use of injectable progestogen contraceptives (IPCs) among breast cancer cases and, controls in greater Cape Town, South Africa, -7 IPC A V l^^\ Oil (*^^ exposure Any use Duration (years) < - - Interval since first use (years) < - - Interval since last use (years) Current userf Cases (n = 7) Controls (n = ) < % Cl.,. Cases Controls (n=) (n = 7) $.7 % Cl /ou Age (years) Cases (n = ).7,. Controls (n = ) % Cl., Cases Controls (n = ) (n=,) * Relative risks and % CIs are adjusted for age (continuous), ethnic group, socioeconomic status, and any combined estrogen/progestogen oral contraceptive (COC) use. t Exposed < year previously. t % Cl does not include , Total..... % Cl.7,.,.7,..,.7,..7,..7,.,.7.7,,.*.7,..,.7,.., Downloaded from by guest on October

5 Shapiro et al. them, and 7, respectively, were exposed to IPCs ( =., percent CI,.). The stage () of the breast cancer was known for 77 cases: there were stage cases, and the relative risk estimate was ; the corresponding numbers (and estimates) for stages,, and were (), (), and (.), respectively. Combined estrogen/progestogen oral contraceptives Among the cases and, controls ( percent) and ( percent), respectively, had used COCs ( =, percent CI.,.) (table ). For total durations of use that extended from < year to > years, the overall relative risk estimates ranged from to., and there was no evidence of a trend. For intervals since first exposure that extended up to or more years, the overall estimates ranged from to.7, and all percent confidence intervals included unity. For current COC use, the relative risk was ( percent CI.,.), and for last exposure \-A years previously, it was. ( percent CI,.); for intervals since last exposure of -, - and > years, the estimates were.,., and, respectively, and compatible with unity. Among 7 cases and controls below age years (table ), the overall relative risk estimate for COC users was.7 ( percent CI.,.). For total durations of exposure of <, -, and - years the relative risks were.,., and., respectively; the former two estimates were statistically significant. For COC use that had commenced <, -, and - years previously, the relative risk estimates were.,. (significant), and.. For current COC use, and for last exposure -, -, and - years previously, the relative risk estimates were.,. (significant),.7, and. (significant), respectively. Among women in age groups - and - years (table ), the overall relative risk estimates were and, respectively, and compatible with unity. There were no significant associations according to duration of use, or according to the interval since first or last use, and no evidence of a declining risk as the interval since last exposure increased. Among and, Colored cases and controls, and, respectively, had used COCs ( =, percent CI,.). Among and Black cases and controls, the corresponding numbers were and 7 ( =.7, percent CI.,.). Among cases and controls who were never exposed to IPCs, and, respectively, had used COCs ( =., percent CI,.). First exposure to COCs before age years was reported by cases and controls ( =, percent CI.,.). Among cases and, controls who were not members of medical aid societies, 7 and, respectively, had used COCs ( =, percent CI,.). Among cases and, controls who had never had a breast examination or mammogram, and, respectively, had used COCs ( =., percent CI,.). For women with stages - breast cancer, the relative risk estimates ( percent CIs) were.7 (,.),. (.,.),, and., respectively. DISCUSSION The present findings suggest that IPCs, principally DMPA, do not increase the overall risk of breast cancer: the relative risk was, and a value of more than could be ruled out with percent confidence. Nor was there evidence to suggest that the risk increased with prolonged duration of use. In the overall data, the only statistically significant relative risk was an estimate of. for current IPC use. However, in the immediately contiguous category of women who had stopped using IPCs - years previously, the estimate was. In addition, the association with current exposure was inconsistent according to age, and it was mainly accounted for by a relative risk estimate of. among women aged ^ years. There was also no evidence to suggest that IPCs cause a transitory increase in the risk during a period lasting - years after exposure has been discontinued, or that their use is associated with the occurrence of breast cancer at a young age, or that IPCs increase the risk after prolonged time lapses, extending to more than two decades. Nor was there evidence to suggest that women who were exposed during adolescence or early adulthood, when the breast was growing rapidly (7), were at increased risk. For women exposed to COCs, there was an overall -fold increase in the risk that was of borderline statistical significance. The risk did not vary significantly according to duration of use, or according to the interval since first or last use. In particular, there was no evidence to suggest that the risk was highest among currently exposed women, followed by a decline as the interval after stopping the use of COCs extended for a decade or longer (, ). The overall relative risk of was mainly accounted for by an estimate of.7 among women less than years of age. In that age group, the risks were uniformly elevated for all durations of use. They were also elevated for intervals since first or last use that extended up to years. Again, there was no evidence to suggest an initial increase in the risk among currently exposed women, followed by a decline, as the interval after stopping COC use increased. Among women in the age groups Downloaded from by guest on October Am J Epidemiol Vol., No.,

6 r TABLE. Relative risks* () (% confidence intervals (Cl)) for the use of combined estrogen/progestogen oral contraceptives (COCs) among breast cancer Z cases and, controls in greater Cape Town, South Africa, -7 ro o COC exposure Any use Duration (years) < - - Interval since first use (years) < - - Interval since last use (years) Current usert : Cases Controls (n = 7) (n = ) Age (years) < -^ - Total.7.*.*...*...*.7.* % Cl Cases Controls (n=) (n = 7).,. 7,., % Cl Cases Controls (n = ) (n = ) * Relative risks and % CIs are adjusted for age (continuous), ethnic group, socioeconomic status, and any injectable progestogen oral contraceptive (IPC) use. t Exposed < year previously. * % Cl does not include % Cl Cases Controls (n = ) (n=,) % Cl.,.,..,.,..7,..7,..,.7,.,..,.,.*,..,..7, Downloaded from by guest on October

7 Shapiro et al. - and - years, all relative risk estimates were compatible with unity. It is important to consider whether bias could have accounted for any of the associations identified in this study. Information bias could have occurred if the cases reported their contraceptive exposures more fully than the controls. For COC users, it is possible that greater awareness of the hypothesis among women below age years may have contributed to the observed association. The selection of the cases could have been biased if breast cancer tended to be diagnosed more commonly, or earlier, among exposed than among nonexposed women (). Evidence to support that possibility was the decline in the relative risk for COC use, from a statistically significant value of.7 among women with stage to. among women with stage breast cancer. Other sources of biased case selection are unlikely because until virtually all cases in the study base were enrolled, and the findings were unchanged when the analysis was confined to women who did not have access to private clinics. It is unlikely that selection bias affected the controls: women were only included if their reason for hospitalization was independent of contraceptive use, and of breast cancer risk, and the enrollment of eligible subjects was virtually percent. The similar usage rates of IPCs and COCs in the major diagnostic categories that comprised the control series is additional evidence against bias. Potential confounding was controlled, partly by series-matching, and more fully by multivariate adjustment. In addition, as can readily be calculated, all crude relative risk estimates approximated the corresponding estimates derived by multiple logistic regression, suggesting that there was no material confounding. Earlier studies of contraceptives that contain progestogens, but no estrogens, have in general suggested that these agents do not increase the overall risk of breast cancer (, 7, ). However, they have been discordant with regard to effects within subgroups. In a case-control study confined to British women below age years (), there was a reduced risk of breast cancer among women who used oral progestogen-only contraceptives for > years. By contrast, two casecontrol studies of DMPA users, one from New Zealand () and the other conducted in several countries by the World Health Organization (WHO) (7) were null overall, but each identified increased risks among women who were less than years of age, and among women who had been exposed < years previously. Exposure at a young age and recent exposure were highly correlated, and in an attempt to distinguish between their separate effects, the investigators pooled the data from the two studies (). They concluded that the probable key factor was an increased risk during the -year period following exposure to DMPA, regardless of age, rather than an elevated risk confined to young women. In, the New Zealand data were again analyzed, this time to assess the effects of oral progestogen contraceptives (). There was no overall increase in breast cancer risk, but increased risks were observed among women below age years, and among those exposed < years previously. Among those exposed > years previously, the risk was significantly reduced. Based on the New Zealand () and WHO (7) studies of DMPA, and on the New Zealand findings for oral progestogen contraceptives (), it has been suggested that exposure to progestogen contraceptives may mimic pregnancy, and have analogous effects on the general pattern of breast cancer risk (). That is, as has been postulated for pregnancy, exposure to progestogens may initially increase the risk for a few years, followed by a decline to an eventual longlasting reduction in the risk. Thus, because breast cancer becomes more common with advancing age, the net effect may be to reduce the lifetime risk. The present findings do not support that suggested pattern of events. In addition, the postulated effect of pregnancy itself on breast cancer risk was not evident in the New Zealand data (), and it is controversial (). For COC use, the present findings are discordant with those reported in a pooled reanalysis of,7 cases and, controls from studies (, ), in which the relative risk estimates for current users, and those who stopped using COCs -, -, and > years previously were,,.7 and., respectively a pattern that was evident in all age groups. In the present study, there was no evidence of a corresponding pattern of initial risk elevation, followed by a decline. Moreover, the positive association with COC use observed in this study was largely confined to the youngest women, among whom the relative risk estimates were in the range of.-. in all categories of recency of use extending up to years previously. In some of the studies included in the pooled analysis, the association was strongest at the youngest ages (,,, ), in accord with the present findings. It is also possible that there may have been shared biases among the included studies that could have accounted for the small risk increments that were observed. In addition, the positive associations were largely confined to early-stage breast cancer, as in the present study. The investigators were unable to conclude whether the findings reflected cause and effect, or bias. It remains to consider the public health implications of the present findings. For women who use IPCs, Am J Epidemiol Vol., No., Downloaded from by guest on October

8 Hormonal Contraception and Breast Cancer DMPA in particular, there was no evidence of an overall increase in the risk of breast cancer, and for currently exposed women, the relative risk was.. With regard to COCs, the overall relative risk was, and it was.7 among women below age years. These associations could have been due to chance or bias. If causation is assumed, however, based on an annual incidence of breast cancer of per, among women who are - years of age, the estimated excess risk (7, ) attributable to current IPC use is per,; for any COC use, it is per,. Among women below age years, based on an annual incidence of per,, the excess risk attributable to any COC use is per,. Thus, even under the most adverse assumptions, the incidence rates of breast cancer among IPC and COC users that can be attributed to exposure are small. ACKNOWLEDGMENTS This work was supported by the US National Cancer Institute (grant no. R CA). The investigators wish to acknowledge the help of Dr. Esther Sapire for providing information on contraceptive use in family planning clinics. Drs. David Dent, Anne Gudgeon, and Esther Sapire served on an Advisory Committee to the study. We are grateful to Carol Richards, Bev Arendse, Phoebe Gribble, and Lungisa Makayi who carried out the data collection and other tasks. We thank the superintendents and staff of the Groote Schuur Hospital, Observatory, Cape, and the Tygerberg Hospital, Belleville, and the medical administration of the Western Cape Province, for their help. REFERENCES. Bhopal R, Donaldson L. Commentary. White, European, Western, Caucasian, or what? Inappropriate labeling in research on race, ethnicity and health. Am J Public Health ;:-7.. Fullilove MT. Comment: abandoning race as a variable in public health research an idea whose time has come. Am J Public Health ;:7-.. Rosenfield A, Maine D, Rochat R, et al. The Food and Drug Administration and medroxyprogesterone acetate. JAMA ; :-.. Weisz J, Ross GT, Stolley PD. Report of the public board of inquiry on Depo-Provera. Rockville, MD: United States Food and Drug Administration,.. Stone R. Depo-Provera. Controversial contraceptive wins approval from FDA panel. Science ;:.. Paul C, Skegg DCG, Spears GFS. Depot medroxyprogesterone (Depo-Provera) and risk of breast cancer. BMJ ;: WHO Collaborative Study of Neoplasia and Steroid Contraceptives. Breast cancer and depot-medroxyprogesterone acetate: a multinational study. Lancet,:-.. Malone KE, Daling JR, Weiss NS. Oral contraceptives in relation to breast cancer. Epidemiol Rev ; :-7.. Romieu I, Berlin JA, Colditz G. Oral contraceptives and breast cancer: review and meta-analysis. Cancer ;:-.. Malone KE. Oral contraceptives and breast cancer: a review of the epidemiological evidence with an emphasis on younger women. In: Committee on the Relationship between Oral Contraceptives and Breast Cancer, Institute of Medicine, eds. Oral contraceptives and breast cancer. Washington, DC: National Academy Press, National Academy of Sciences, :7-.. Thomas DB. Oral contraceptives and breast cancer: review of the epidemiologic literature. Contraception ;:7-.. Prentice RL, Thomas DB. On the epidemiology of oral contraceptives and disease. Adv Cancer Res 7;:-.. Collaborative Group on Hormonal Factors on Breast Cancer. Breast cancer and hormonal contraceptives. Lancet ;7: Collaborative Group on Hormonal Factors on Breast Cancer. Breast cancer and hormonal contraceptives. Contraception ;(suppl):ls-s.. Bailie R, Katzenellenbogen J, Hoffman M, et al. A case control study of breast cancer risk and exposure to injectable progestogen contraceptives. S Afr Med J 7;7:-.. Central Statistical Service. Population Census. Report No. --. Pretoria, South Africa: Central Statistical Service,. 7. Pike MC, Henderson BE, Krailo MD, et al. Breast cancer in young women and use of oral contraceptives: possible modifying effect of formulation and age at use. Lancet ;: -.. Skegg DCG. Potential for bias in case-control studies of oral contraceptives and breast cancer. Am J Epidemiol ;7: -.. Hermanek P, Sobin LH, eds. TNM classification of malignant tumours. th ed. Berlin, Germany: Springer-Verlag,.. Staffa JA, Mewschaffer CJ, Jones JK, et al. Progestins and breast cancer: an epidemiologic review. Fertil Steril ;7: 7-.. UK National Case-Control Study Group. Oral contraceptive use and breast cancer risk in young women. Lancet ;: 7-.. Skegg DCG, Noonan EA, Paul C, et al. Depot medroxyprogesterone acetate and breast cancer. A pooled analysis of the World Health Organization and New Zealand studies. JAMA ;7:7-.. Skegg DCG, Paul C, Spears GFS, et al. Progestogen-only oral contraceptives and risk of breast cancer in New Zealand. Cancer Causes Control ;7:-.. Kelsey JL, Gammon MD, John EM. Reproductive factors and breast cancer. Epidemiol Rev ; :-7.. White E, Malone KE, Weiss NS, et al. Breast cancer among young US women in relation to oral contraceptive use. J Natl Cancer Inst ;:-.. Rookus MA, van Leeuwen FE. Oral contraceptive and risk of breast cancer in women aged - years. Netherlands Oral Contraceptives and Breast Cancer Study Group. Lancet ; :-. 7. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. Belmont, CA: Lifetime Learning Publications, :-, -7.. Miettinen OS. Proportion of disease caused or prevented by a given exposure, trait or intervention. Am J Epidemiol 7;:-. Downloaded from by guest on October Am J Epidemiol Vol., No.,